The autophagy GABARAPL1 gene is epigenetically regulated in breast cancer models

GABARAP family genes are differentially expressed in human breast cancer biopsies

We first analyzed GABARAP, GABARAPL1 or GABARAPL2 mRNA expression in human BC biopsies using qRT-PCR (Fig. 1). The different BC subtypes are classified in regard of their molecular marker expression.
Luminal BC, which represent 50 % of total BC and generally associated with a good
prognosis, are divided in Luminal A and Luminal B BC. Both Luminal A and B BC express
ER? (estrogen receptor: ER+) whereas the expression of the HER2 (HER+) (human epidermal growth factor receptor) gene was only observed in Luminal
B BC. HER+ BC subtype, which represents 17 % of BC, presents an amplification of the
HER2 gene without the expression of ER? (ER-). The triple negative BC (ER-/PR-/HER-) do
not express ER?, PR (progesterone receptor) and HER2 and cannot be treated with specific
therapies (for a review, see 24]). Our cohort comprised 5 grade I BC (ER+/PR+/HER-) and 8 grade III BC (5 ER-/PR-/HER-and
3 ER+/PR+/HER+). Our results revealed an insignificant decrease of both GABARAP and GABARAPL2 mRNA levels in grade III BC compared to non tumoral tissue (NT). More interestingly,
GABARAPL1 expression was strongly decreased in BC grade III tissues (p?=?0.004) versus non tumoral tissues. An inverse correlation (r?=??0.57) was also observed between GABARAPL1 mRNA and tumor stage while a very poor correlation was determined between GABARAP or GABARAPL2 expression levels and the tumor grade (respectively r?=??0.3 and r?=??0.1).

Fig. 1. GABARAP family expression is deregulated in breast cancers. Quantification of GABARAP, GABARAPL1, GABARAPL2 expression using qRT-PCR, in grade I?+?II and grade III BC biopsies compared to normal
adjacent tissue. Circle: NT (non tumoral), square: grade II, triangle: grade III.
Difference of expression were quantified using a t-test. The correlation between the
tumor grade and gene expression was measured using a Spearman test

Study of global and local epigenetic modifications in GABARAPL1 promoter and GABARAPL1 expression in BC tissues and cell models

Since tumors are frequently associated with aberrant DNA methylation content, we quantified
global DNA methylation levels using ELISA in gDNA issued from both grade III BC samples
(associated with the lowest expression of GABARAPL1 mRNA, as previously described in Fig. 1 and NT tissues (Fig. 2a).
⁷25], 26] As expected, several gDNA issued from grade III BC presented a lower global DNA methylation
compared to gDNA issued from NT tissues. These results suggested that epigenetic modifications
might occur in these tumors. To determine whether local DNA methylation was also altered
in these BC samples, methylation of XIST, a X-linked gene known to be methylated in women, and GAPDH, a gene constitutively active and unmethylated, were analyzed by precipitation of
methylated DNA using the methylCollector Ultra kit. As expected, methylation of XIST was observed in 100 % of both gDNA issued from NT and grade I-II BC tissues tested
but lost in 3 out of 5 triple negative BC biopsies (Grade III) (Additional file 1: Figure S1). On the contrary, GAPDH was never found to be methylated in NT gDNA but was surprisingly frequently methylated
in gDNA issued from BC samples (Additional file 1: Figure S1). Loss of global DNA methylation in several tissues (Fig. 2a 4 out 13 samples) and aberrant status of methylation of XIST and GAPDH (Additional file 1: Figure S1) strongly suggested that DNA methylation was deregulated in BC samples.

Fig. 2. Detection of DNA methylation in GABARAP family promoters a Global DNA methylation quantification in BC and non tumoral biopsies using methylFlash
ELISA kit. b Scheme describing the GABARAP family gene promoter structure (Methprimer) and primer localization. c Left: Descriptive of GABARAP family gene methylation using methylCollector kit ; right : examples of methylation
signal observed using GABARAPL1 MC2 primers. White : absence of signal of methylation; black : signal of methylation.
d Correlations (Spearman test) between expression and methylation of GABARAP, GABARAPL1 and GABARAPL2 genes. NT: non tumoral, ER +/?: status of expression of estrogen receptor ?, PR+/?:
status of expression of progesterone receptor, HER+/?: status of expression of Human
epidermal growth factor receptor; I-III: BC grade

We next analyzed whether epigenetic regulation of the GABARAPL1 might explain its specific down-regulation in BC. GABARAP, GABARAPL1 and GABARAPL2 present CpG rich areas in their promoter (?800/+200) as predicted using the methPrimer
software 27], so we analyzed the methylation status of these promoters using the methylCollector
Ultra kit (Fig. 2b and c). A low methylation signal was observed in both GABARAP and GABARAPL2 promoters (MC primers) both in tumors and NT tissues while primers designed in the–800
region of GABARAPL1 (MC1) (Fig. 2c) revealed a strong signal of methylation in both NT and BC tissues. Regarding GABARAPL1, a low signal of unmethylation was also measured in NT samples but not in BC samples
suggesting that hemi-methylation was lost in cancer cells (p?=?0.0024) (Fig. 2c). To confirm the higher level of methylation in the GABARAPL1 promoter in BC, the same experiment was repeated using primers (MC2) designed to
detect the 5?-UTR region of this gene. Results obtained with MC2 primers showed that
GABARAPL1 was not or weakly methylated in NT samples but highly methylated in BC tissues (p?=?0.05) (Fig. 2c). Pearson correlation analysis revealed that GABARAP or GABARAPL2 expression was not correlated with methylation status (r?=??0.45, p?=?0.12 and r?=?0.36, p?=?0.25 Fig. 2d), while GABARAPL1 expression was indeed correlated with the methylation status of the gene (r?=??0.58,
p?=?0.03 suggesting that methylation of the GABARAPL1 promoter may explain the downregulation of GABARAPL1 expression in BC (Fig. 2d).

In order to characterize the pathway allowing epigenetic modifications to control
GABARAP family expression, we first analyzed GABARAP, GABARAPL1 and GABARAPL2 mRNA levels using qRT-PCR in MCF-7 and MDA-MB-453 (BC cell lines) and MCF-10A (Breast
immortalized but non tumoral cell line) cell lines (Fig. 3a). Both GABARAP (p?=?0.039), GABARAPL1 (p?=?0.013) and GABARAPL2 (p?=?0.039) expression were decreased in MCF-7 compared to MCF-10A but while the loss
of GABARAP or GABARAPL2 expression was weak (0.5-2 fold), GABARAPL1 expression was about 100-fold lower in MCF-7 compared to MCF-10A (Fig. 3a). Similar results were obtained for GABARAP, GABARAPL1 and GABARAPL2 in MDA-MB-453 cells but GABARAP did not show any significant differences in this cell line (Fig. 3a). Methylation of the GABARAP family promoters was then assessed using the MethylCollector kit in both MCF-7 and
MDA-MB-453 cell lines presenting a low expression of GABARAPL1. First, a signal of methylation of GABARAP promoter was observed in MCF-7 cells but not in MDA-MB-453 cells (Fig. 3b). Regarding GABARAPL2 promoter, no methylation was detected in both cell lines. A high signal of methylation
was detected in the GABARAPL1 promoter in MCF-7 and MDA-MB-453 cells using MC1 primers confirming that the region
of MC1 is highly frequently methylated. As expected GABARAPL1 promoter methylation was lost in MCF-7 cells treated with 5-aza-CdRdeoxycytidine
(5-aza-CdR), a DNMTi (Fig. 3b). A high signal of methylation was also detected using MC2 primers in MCF-7 cells
but not in MCF-10A cells suggesting, as observed before in human BC biopsies, that
GABARAPL1 methylation is predominantly observed in BC cell lines (Fig. 3c)

Fig. 3. Epigenetic modifications in GABARAP family gene promoters. a Quantification of GABARAP, GABARAPL1, GABARAPL2 expression using qRT-PCR in BC MCF-7 and MDA-MB-453 cancer cells and MCF-10A immortalized
cells. b and cGABARAP family gene methylation using methylCollector kit in MCF-7, MDA-MB-453 and MCF-10A
cells. (I: input; M: methylated fraction). d Visualization of H3 deacetylation using ChIP experiment and anti-H3 acetylated (H3-ac)
antibody in the GABARAP family gene in MCF-7 and MDA-MB-453 cells (I: input; IgG: negative control of IP).
e Detection of DNMT1 and HDAC1 recruitment on GABARAPL1 promoter using ChIP experiment and anti-DNMT1 or anti-HDAC1 antibody (I: input; IgG
: negative control)

As local DNA methylation is frequently associated to histone deacetylation, the acetylation
status of histone H3 (H3-ac) was analyzed by ChIP in MCF-7 and MDA-MB-453 cells previously
treated or not with trichostatin A (TSA), an inhibitor of HDACs (Fig. 3d). ChIP analysis revealed that H3 acetylation was detected in the GABARAP promoter in MCF-7 cells but not in MDA-MB-453 cells. A very low level of H3 acetylation
was also observed in GABARAPL2 promoter in these both cell lines. These signals were increased following TSA treatment
in MCF-7 and MDA-MB-453 cells. Similarly, no H3 acetylation signal could be detected
in the GABARAPL1 promoter of BC cells but this signal was increased after TSA treatment, particularly
in the MDA-MB-453 cells (Fig. 3d).

All these data (Figs. 1, 2 and 3) suggest that GABARAPL1 is the most regulated gene of the GABARAP family and that require promoter deacetylation and DNA methylation. Interestingly,
both DNMT1, which predominantly catalyzes inheritance DNA methylation, and HDAC1 were
detected on GABARAPL1 promoter in MCF-7 cells using ChIP experiments (Fig. 3e).

We next asked whether 5-aza-CdR or TSA could restore GABARAPL1 expression in BC cell lines. To do so, MCF-7 and MDA-MB-453 cells were treated with
5-aza-CdR or TSA and the levels of GABARAP, GABARAPL1 and GABARAPL2 mRNA were measured using qRT-PCR (Fig. 4a) while protein levels were quantified by western-blotting (Fig. 4b). All cells were also treated with MG-132, before protein extraction, to prevent
the fast proteasomal degradation of GABARAPL1 which has been previously reported 28]. First we observed a not significant increase of GABARAP mRNA (p?=?0.07) but a significant increase of the corresponding protein GABARAP (p?=?0.007 and p?=?0.02) following TSA treatment in MCF-7 and MDA-MB-453 cells. Both GABARAPL2 mRNA (p?=?0.05 and p?=?0.05 respectively) and GABARAPL2 protein (p?=?0.045 and p?=?0.0003 respectively) were increased in MCF-7 and MDA-MB-453 cells treated with
TSA. No significant effect of 5-aza-CdR could be observed on GABARAP and GABARAPL2 expression (Fig. 4a). On the opposite, 5-aza-CdR treatment increased GABARAPL1 mRNA level (about 2 fold) in both cell lines (p?=?0.004 and p??0.0001 respectively) while TSA treatment increased GABARAPL1 expression of 10 to 25 fold (Fig. 4a). Moreover, GABARAPL1 expression, which was undetectable in non treated MCF-7 cells
or 5-aza-CdR-treated cells, was increased following TSA treatment (Fig. 4b). A weak and diffuse band signal corresponding to the GABARAPL1 protein was also
in over-exposed western-blotting using lysates of MDA-MB-453 cells treated with TSA,
suggesting that GABARAPL1 might also be slightly increased in these cells after TSA
treatment.

Fig. 4. Effects of epigenetic modulators on the GABARAP family gene expression. a Effects of 5-aza-deoxycytidine (5-aza-CdR) or trichostatin A (TSA) on GABARAP family gene expression using qRT-PCR analysis. b Effects of 5-aza-CdR or TSA on GABARAP family protein expression using WB (anti-GABARAP/GABARAPL1,
anti-GABARAPL2 and anti-ACTIN antibodies) in cells previously treated with MG-132.
Differences were quantified using a t-test

Altogether these results confirm that GABARAPL1 is the gene of the GABARAP family whose expression is the most sensitive to epigenetic regulation in BC cell
lines. Since 5-aza-CdR and TSA treatments restored GABARAPL1 content, we next asked
whether these compounds modulate autophagy and cell proliferation in MCF-7 cells (Additional
file 2: Figure S2). Both an increase of LC3B-II (form associated to the autophagosomes)
(Additional file 2: Figure S2A) and of cells with GFP-LC3 puncta (Additional file 2: Figure S2B) were observed in respectively 5-aza-CdR/ TSA treated cells and in GFP-LC3
transfected and TSA treated cells compared to control cells. Moreover, both an decrease
of cell proliferation and clonogenecity were also observed in MCF-7 treated cells
(Additional file 2: Figure S2C and D). These results suggest that restoration of GABARAPL1 expression
might be linked to these processes although some pleiotropic effects of 5-aza-CdR
and TSA treatment could also be involved 29].

GABARAPL1 expression is regulated by CREB-1

Behind epigenetic modifications of the promoter, the second step of gene regulation
is the recruitment of transcriptional factors (TF). Therefore, in order to characterize
the mechanisms governing GABARAPL1 expression in these cells, we cloned two fragments of the GABARAPL1 promoter (?659/+241 and ?336/+241) in the pGL3 luciferase reporter plasmid. As suggested
by 3 different softwares (TESS http://www.cbil.upenn.edu/tess, Cister http://zlab.bu.edu/~mfrith/cister.shtml and Patch http://www.gene-regulation.com/pub/programs.html#patch) used to predict TF binding sites, several putative CRE (cAMP Response Element) elements
were identified in GABARAPL1 promoter (Fig. 5a). Indeed, following transfection of our constructions in MCF-7 cells, we observed
a significant increase of luciferase signal in cells transfected with the–336/+241-GABARAPL1-promoter-pGL3 plasmid compared to the empty pGL3 vector suggesting the presence of
functional regulatory elements in this region. Moreover, luciferase activity was strongly
increased when cells were transfected with the–659/+241-GABARAPL1–promoter-pGL3 vector compared to the basal–336/+241-GABARAPL1-promoter-pGL3 vector suggesting that the region ?659/-336 is also important for GABARAPL1 regulation (Fig. 5b). Both treatment of MCF-7 cells with forskolin, a compound known to activate CREB-1
(CRE binding protein-1), or transfection with a plasmid expressing the CREB-1 protein
significantly increased luciferase activity linked to the construction–659/+241-GABARAPL1–promoter-pGL3, suggesting that CREB-1 is indeed involved in GABARAPL1 expression (Fig. 5c). The increase of luciferase activity in cells transfected with the pcDNA3.1-CREB-1
vector might be explained by the low level of endogenous CREB-1 in MCF-7 cells as
observed in IF experiments (Fig. 5c) and ChIP experiment also confirmed the recruitment of CREB-1 on the–659/+241-GABARAPL1-promoter in spite of the presence of a high background noise that may be provoked
by the presence of–659/+241-GABARAPL1-promoter plasmid (Fig. 5d). We next asked whether endogenous GABARAPL1 expression may be regulated by CREB-1. Treatment of MCF-7 cells with Actinomycin
D revealed that the half-life of GABARAPL1 mRNA was high (about 17 h) suggesting that regulation of GABARAPL1 mRNA content is more dependent on transcription that mechanisms affecting mRNA stability
(Fig. 5e) 30], 31]. Moreover, overexpression of CREB-1 in MCF-7 cells significantly increased GABARAPL1 expression but at a lower level than the ones observed in cells treated with 5-aza-CdR/TSA
(Fig. 5f). Moreover, a non-significant (p?=?0.066) further increase of GABARAPL1 expression was observed in cells transfected with the CREB-1 plasmid and treated
with 5-aza-CdR/TSA compared to cells with 5-aza-CdR/TSA treatment alone.

Fig. 5. GABARAPL1 expression I controlled by CREB-1. a Scheme describing the position of primers used (in regard of the putative initial
transcription site (+1)) and putative CRE (CREB-1 response elements) sites in the
GABARAPL1 promoter. b and c Luciferase activity measured using a Luciferase assay System Kit in MCF-7 cells transfected
with empty pGL3 plasmid,?336/+241-GABARAPL1-promoter-pGL3 plasmid,–659/+241-GABARAPL1-promoter-pGL3 plasmid, pCDNA3.1-CREB-1 or treated with (10 ?M) forskolin. c Bottom : expression of CREB-1 using IF in cells transfected or not with the pCDN3A.1-CREB-1
vector. d Recruitment of CREB-1 on–659/+241-GABARAPL1-promoter-pGL3 plasmid using ChIP experiment and anti-CREB-1 antibody in MCF-7 cells
transfected with–659/+241-GABARAPL1-promoter-pGL3 and pCDNA3.1-CREB-1 plasmids (I: input; IgG : negative control of IP).
e Half-life of GABARAPL1 mRNA using qRT-PCR following Actinomycin D treatment in MCF-7 cells. f Effects of CREB-1 overexpression (following pCDNA3.1-CREB-1 plasmid transfection)
and/or 5-aza-CdR /TSA treatment on GABARAPL1 expression using qRT-PCR in MCF-7 cells. g Effects of CREB-1 overexpression (following pCDNA3.1-CREB-1 plasmid transfection)
and/or 5-aza-CdR/TSA treatment on CREB-1 recruitment in GABARAPL1 promoter using ChIP experiment and an anti-CREB-1 antibody (I: input; IgG : negative
control). Differences were quantified using a t-tests. GL1: GABARAPL1

These data suggest that epigenetic modifications may be an initial predominant factor
allowing the recruitment of CREB-1 on GABARAPL1 promoter. These observations were partly confirmed by ChIP experiments showing an
increase of CREB-1 recruitment on GABARAPL1 promoter in MCF-7 cells previously treated with 5-aza-CdR/TSA. But no significant
difference could be observed between 5-aza-CdR/TSA treated cells transfected or not
with the vector encoding CREB-1 (Fig. 5g).